Joint Power Optimization and AP Selection for Secure Cell-Free Massive MIMO
Yasseen Sadoon Atiya, Zahra Mobini, Hien Quoc Ngo, Michail Matthaiou
TL;DR
The paper addresses secrecy in CF-mMIMO under active eavesdropping by jointly optimizing AP selection and downlink power control to minimize the eavesdropper's spectral efficiency while meeting legitimate users’ QoS. It converts the challenging mixed-integer nonconvex problem into a tractable form and solves it with an accelerated projected gradient (APG) method enhanced by a penalty framework and variable reparameterization. The proposed approach yields substantial secrecy performance gains, outperforming heuristic schemes (up to 265% higher 50th percentile SSE) and showing larger advantages when Eve is closer to the targeted user. This work demonstrates a scalable, low-complexity pathway to bolster physical-layer security in CF-mMIMO systems with active attacks, enabling more robust secure communications in dense deployments.
Abstract
In this paper, we investigate joint power control and access point (AP) selection scheme in a cell-free massive multiple-input multiple-output (CF-mMIMO) system under an active eavesdropping attack, where an eavesdropper tries to overhear the signal sent to one of the legitimate users by contaminating the uplink channel estimation. We formulate a joint optimization problem to minimize the eavesdropping spectral efficiency (SE) while guaranteeing a given SE requirement at legitimate users. The challenging formulated problem is converted into a more tractable form and an efficient low-complexity accelerated projected gradient (APG)-based approach is proposed to solve it. Our findings reveal that the proposed joint optimization approach significantly outperforms the heuristic approaches in terms of secrecy SE (SSE). For instance, the $50\%$ likely SSE performance of the proposed approach is $265\%$ higher than that of equal power allocation and random AP selection scheme.